all_trials %>% 
    count(pilot, response_type) %>% 
    pivot_wider(names_from=response_type, values_from=n) %>% 
    replace(is.na(.), 0) %>% 
    pivot_longer(!c(pilot, pilot), names_to="response_type", values_to="n", names_prefix="") %>%
    group_by(pilot) %>% mutate(prop = n/sum(n)) %>% ungroup() %>% 
    ggplot(aes(pilot, prop, fill=fct_rev(response_type))) +
    geom_bar(stat="identity") +
    scale_colour_manual(values=c(
        "deeppink2",
        "deeppink4", 
        "gray", 
        "springgreen4"
    ), aesthetics=c("fill", "colour"), name="") +
    labs(x="", y="Proportion of Trials") + coord_flip()

response_colors = scale_colour_manual(values=list(
    other="deeppink2",
    intrusion="deeppink4", 
    empty="gray", 
    correct="springgreen4"
), aesthetics=c("fill", "colour"), name="")

all_trials %>% 
    count(pilot, wid, response_type) %>% 
    pivot_wider(names_from=response_type, values_from=n) %>% 
    replace(is.na(.), 0) %>% 
    mutate(wid=fct_reorder(wid, 100*correct + empty)) %>% 
    pivot_longer(!c(pilot, wid), names_to="response_type", values_to="n", names_prefix="") %>%
    group_by(pilot, wid) %>% mutate(prop = n/sum(n)) %>% ungroup() %>% 
    ggplot(aes(wid, prop, fill=fct_rev(response_type))) +
    geom_bar(stat="identity") +
    response_colors +
    labs(x="Participant", y="Proportion of Trials") + 
    scale_x_discrete(breaks=NULL) +
    facet_wrap(~pilot, scales="free_x")

all_trials$name = 'Human'

excl = all_trials %>% 
    group_by(pilot,wid) %>%
    summarise(n_skip = sum(skip)) %>% 
    mutate(keep=between(n_skip, 4, 33))

excl %>%
    mutate(excluded = if_else(keep, "included", "excluded")) %>% 
    ggplot(aes(pilot, fill=excluded)) + 
    geom_bar() +
    ylab("Number of Participants") +
    scale_colour_manual(values=c(
        "gray", "dodgerblue"
    ), aesthetics=c("fill", "colour"), name="") + coord_flip() + xlab("")

keep = excl %>% filter(keep) %>% with(wid)

trials = all_trials %>% 
    filter(wid %in% keep) %>% 
    group_by(wid) %>% 
    mutate(rt_z = zscore(rt)) %>% 
    ungroup()

pretest = pretest %>% filter(wid %in% keep)

pretest %>% 
    filter(block == max(block)) %>% 
    rename(pre_correct = correct) %>% 
    mutate(pre_logrt = if_else(pre_correct, log(rt), 0)) %>% 
    group_by(wid, word) %>% 
    summarise(across(c(pre_correct, pre_logrt), mean)) %>%
    group_by(wid) %>%
    mutate(across(c(pre_correct, pre_logrt), zscore, .names="{.col}_z")) %>% 
    mutate(
        raw_strength = -((1-pre_correct) * log(15000) + pre_logrt),
        strength = zscore(raw_strength)
    ) %>% 
    right_join(trials) -> trials

trials %>% #plot
    filter(response_type == "correct") %>% 
    regress(judgement, rt) +
    xlab("Confidence Judgement")

trials %>%  #plot
    filter(skip) %>% 
    regress(judgement, rt, bins=0, bin_range=1) +
    stat_summary(fun.data=mean_cl_boot, size=.2) +
    xlab("Feeling of Knowing Judgement")

trials %>% #plot
    filter(skip) %>% 
    regress(strength, rt) + 
    coord_cartesian(xlim=c(NULL), ylim=c(0, 4000))

trials %>% #plot
    filter(skip) %>% 
    group_by(wid) %>% 
    filter(sd(strength) != 0) %>% 
    mutate(strength=zscore(strength), rt_z = zscore(rt)) %>% 
    regress(strength, rt_z) +
    coord_cartesian(xlim=c(NULL), ylim=c(-1.5, 1.5))

trials %>%
    filter(skip) %>% 
    group_by(wid) %>% 
    filter(sd(strength) != 0) %>% 
    mutate(strength=zscore(strength), rt_z = zscore(rt)) %>% 
    group_by(version) %>% 
    group_modify(function(data, grp) {
        lmer(rt_z ~ strength + (strength|wid), data=data) %>% tidy
    }) %>% 
    filter(term == "strength") %>% 
    select(version, estimate, std.error, p.value) %>% 
    left_join(
        load_data('participants') %>% count(version)
    ) %>% kable(digits=4)

trials %>% 
    group_by(wid) %>% 
    filter(skip) %>% 
    ggplot(aes(strength, rt, group=wid)) + 
    geom_smooth(method="lm", level=0, size=.5, color="black") +
    facet_wrap(~pilot) +
    pretty_labs("strength", "rt")

trials %>% 
    group_by(pilot, wid) %>% 
    filter(skip) %>% 
    summarise(x=mean(pre_correct==0))  %>% 
    group_by(pilot) %>% 
    summarise(n=sum(x != 1), prop=mean(x != 1)) %>% kable(digits=2)

trials %>% 
    ggplot(aes(strength, 1*skip, color=pilot)) +
    geom_smooth(method = "glm", method.args = list(family = "binomial"), formula=y~x) +
    pretty_labs("strength", "probability of skipping")

trials %>% 
    ggplot(aes(raw_strength, 1*skip, group=wid)) +
    geom_smooth(method = "glm", method.args = list(family = "binomial"), formula=y~x, 
        se=F, color="black", size=.5) +
    facet_wrap(~pilot) +
    theme(
        panel.grid.major.x = element_line(color="gray"),
        panel.grid.major.y = element_line(color="gray"),
    ) + geom_vline(xintercept=-log(15000), color="red2") +
    pretty_labs("raw strength", "probability of skipping")

pretest %>%
    group_by(pilot, wid) %>% 
    summarise(accuracy=mean(correct)) %>% 
    ggplot(aes(pilot, accuracy)) +
    geom_quasirandom(color="gray") +
    stat_summary(fun.data=mean_cl_boot) +
    coord_flip() +
    xlab("") +
    ylim(0, 1)

trials %>%
    mutate(response_type = case_when(
        response_type == "correct" ~ "correct",
        response_type == "empty" ~ "empty",
        TRUE ~ "error"
    )) %>% 
    ggplot(aes(pilot, rt, color=response_type)) +
    # geom_quasirandom(color="gray") +
    stat_summary(fun.data=mean_cl_boot) +
    coord_flip() +
    scale_colour_manual(values=c(
        "springgreen4",
        "gray", 
        "deeppink3"
    ), aesthetics=c("color"), name="") +
    xlab("")